nudt-compiler-cpp/scripts/test_compiler.sh

#!/usr/bin/env bash
set -euo pipefail

ROOT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")/.." && pwd)"
COMPILER="$ROOT_DIR/build/bin/compiler"
TMP_DIR="$ROOT_DIR/build/test_compiler"
TEST_DIRS=("$ROOT_DIR/test/test_case/functional" "$ROOT_DIR/test/test_case/performance")
CC_BIN="${CC:-cc}"
RUNTIME_SRC="$ROOT_DIR/sylib/sylib.c"
RUNTIME_OBJ="$TMP_DIR/sylib.o"
LLC_BIN="${LLC:-llc}"
CLANG_BIN="${CLANG:-clang}"

if [[ ! -x "$COMPILER" ]]; then
  echo "未找到编译器: $COMPILER"
  echo "请先构建编译器，例如: mkdir -p build && cd build && cmake .. && make -j"
  exit 1
fi

mkdir -p "$TMP_DIR"

if ! command -v "$LLC_BIN" >/dev/null 2>&1; then
  echo "未找到 llc: $LLC_BIN"
  echo "请安装 LLVM，或通过 LLC 环境变量指定 llc 路径"
  exit 1
fi

if ! command -v "$CLANG_BIN" >/dev/null 2>&1; then
  echo "未找到 clang: $CLANG_BIN"
  echo "请安装 Clang，或通过 CLANG 环境变量指定 clang 路径"
  exit 1
fi

# 编译运行库（供链接生成的可执行文件）
runtime_ready=0
if [[ -f "$RUNTIME_SRC" ]]; then
  if "$CC_BIN" -c "$RUNTIME_SRC" -o "$RUNTIME_OBJ" >/dev/null 2>&1; then
    runtime_ready=1
  else
    echo "[WARN] 运行库编译失败，生成的可执行文件将不链接 sylib: $RUNTIME_SRC"
  fi
else
  echo "[WARN] 未找到运行库源码: $RUNTIME_SRC"
fi

ir_total=0
ir_pass=0
result_total=0
result_pass=0

ir_failures=()
result_failures=()

function normalize_file() {
  sed 's/\r$//' "$1"
}

for test_dir in "${TEST_DIRS[@]}"; do
  if [[ ! -d "$test_dir" ]]; then
    echo "跳过不存在的测试目录: $test_dir"
    continue
  fi

  shopt -s nullglob
  for input in "$test_dir"/*.sy; do
    ir_total=$((ir_total+1))
    base=$(basename "$input")
    stem=${base%.sy}
    out_dir="$TMP_DIR/$(basename "$test_dir")"
    mkdir -p "$out_dir"
    ll_file="$out_dir/$stem.ll"
    stdout_file="$out_dir/$stem.stdout"
    expected_file="$test_dir/$stem.out"
    stdin_file="$test_dir/$stem.in"

    echo "[TEST] $input"
    
    # 编译并捕获所有输出
    compiler_status=0
    compiler_output=""
    compiler_output=$("$COMPILER" --emit-ir "$input" 2>&1) || compiler_status=$?
    
    # 临时文件存储原始输出
    raw_ll="$out_dir/$stem.raw.ll"
    printf '%s\n' "$compiler_output" > "$raw_ll"
    
    # 检查编译是否成功
    if [[ $compiler_status -ne 0 ]]; then
      echo "  [IR] 编译失败: 返回码 $compiler_status"
      ir_failures+=("$input: compiler failed ($compiler_status)")
      # 失败：保留原始输出（包含所有调试信息）
      cp "$raw_ll" "$ll_file"
      rm -f "$raw_ll"
      continue
    fi
    
    # 从混杂输出中提取 IR：
    # - 顶层实体：define/declare/@global
    # - 基本块标签
    # - 缩进的指令行
    # - 函数结束花括号
    grep -E '^(define |declare |@|[[:space:]]|})|^[A-Za-z_.$%][A-Za-z0-9_.$%]*:$' "$raw_ll" > "$ll_file"

    # 检查是否生成了有效函数定义
    if ! grep -qE '^define ' "$ll_file"; then
      echo "  [IR] 失败: 未生成有效函数定义"
      ir_failures+=("$input: invalid IR output")
      # 失败：保留原始输出
      cp "$raw_ll" "$ll_file"
      rm -f "$raw_ll"
      continue
    fi

    # 可选：删除多余的空行
    sed -i '/^$/N;/\n$/D' "$ll_file"
    
    rm -f "$raw_ll"
    
    ir_pass=$((ir_pass+1))
    echo "  [IR] 生成成功 (IR已保存到: $ll_file)"
    
    # 运行测试
    # 运行测试部分
    if [[ -f "$expected_file" ]]; then
      result_total=$((result_total+1))
      
      # 运行生成的可执行文件（优先链接运行库）
      run_status=0
      obj_file="$out_dir/$stem.o"
      exe_file="$out_dir/$stem"
      if ! "$LLC_BIN" -filetype=obj "$ll_file" -o "$obj_file" > "$stdout_file" 2>&1; then
        echo "  [RUN] llc 失败"
        result_failures+=("$input: llc failed")
        continue
      fi

      if [[ $runtime_ready -eq 1 ]]; then
        if ! "$CLANG_BIN" "$obj_file" "$RUNTIME_OBJ" -o "$exe_file" >> "$stdout_file" 2>&1; then
          echo "  [RUN] clang 链接失败"
          result_failures+=("$input: clang link failed")
          continue
        fi
      else
        if ! "$CLANG_BIN" "$obj_file" -o "$exe_file" >> "$stdout_file" 2>&1; then
          echo "  [RUN] clang 链接失败"
          result_failures+=("$input: clang link failed")
          continue
        fi
      fi

      if [[ -f "$stdin_file" ]]; then
        "$exe_file" < "$stdin_file" > "$stdout_file" 2>&1 || run_status=$?
      else
        "$exe_file" > "$stdout_file" 2>&1 || run_status=$?
      fi
      
      # 读取预期文件内容
      expected_content=$(normalize_file "$expected_file")
      
      # 判断预期文件是只包含退出码，还是包含输出+退出码
      if [[ "$expected_content" =~ ^[0-9]+$ ]]; then
        # 只包含退出码
        expected=$expected_content
        if [[ "$run_status" -eq "$expected" ]]; then
          result_pass=$((result_pass+1))
          echo "  [RUN] 返回值匹配: $run_status"
          rm -f "$stdout_file"
        else
          echo "  [RUN] 返回值不匹配: got $run_status, expected $expected"
          result_failures+=("$input: exit code mismatch (got $run_status, expected $expected)")
        fi
      else
        # 包含输出和退出码（最后一行是退出码）
        expected_output=$(head -n -1 <<< "$expected_content")
        expected_exit=$(tail -n 1 <<< "$expected_content")
        actual_output=$(cat "$stdout_file")
        
        if [[ "$run_status" -eq "$expected_exit" ]] && [[ "$actual_output" == "$expected_output" ]]; then
          result_pass=$((result_pass+1))
          echo "  [RUN] 成功: 退出码和输出都匹配"
          rm -f "$stdout_file"
        else
          echo "  [RUN] 不匹配: 退出码 got $run_status, expected $expected_exit"
          if [[ "$actual_output" != "$expected_output" ]]; then
            echo "        输出不匹配"
          fi
          result_failures+=("$input: mismatch")
        fi
      fi
    else
      echo "  [RUN] 未找到预期返回值文件 $expected_file，跳过结果验证"
    fi
  done
  shopt -u nullglob
done

# 输出统计
cat <<EOF
测试完成。
IR 生成: $ir_pass / $ir_total
结果匹配: $result_pass / $result_total
EOF

# 输出失败列表
if [[ ${#ir_failures[@]} -gt 0 ]]; then
  echo ""
  echo "IR 失败列表:"
  for item in "${ir_failures[@]}"; do
    echo "  $item"
  done
fi

if [[ ${#result_failures[@]} -gt 0 ]]; then
  echo ""
  echo "结果失败列表:"
  for item in "${result_failures[@]}"; do
    echo "  $item"
  done
  echo ""
  echo "失败的测试文件已保留在: $TMP_DIR"
  echo "可以查看对应的.ll文件进行调试"
fi